Embed Doc
Copy Link
Readcast
Collections

CommentGo Back

Download

A Bi-Directional Interface Between the Human Nervous System and the InternetK. Warwick, P. Kyberd, M. Gasson, B. Hutt and I. GoodhewDepartment of Cybernetics, University of Reading, UK.Abstract

This paper describes an investigation into the use of implant technology in forming a bi-directional link between the human nervous system and a computer. The microelectrodeimplant and the required neurosurgery are described in detail. Various applicationstudies are described, including interaction with an articulated robot hand, a group of autonomous mobile robots and extra sensory (ultrasonic) signal inputs to the humannervous system. Conclusions are drawn, as a result of the study, as to the long termimplant possibilities and future capabilities of those fitted with implants.

Introduction

Interaction between humans and computers is almost exclusively brought about by meansof signals external to the human body [1], despite the fact that this causes time delays,errors and communication problems. The range of interaction tools is broad, includingkeyboard, joystick, voice activation and visual stimuli. In each case different translationand timing problems occur. The main problem is that electro-chemical signals on thehuman nervous system need to be translated into some form of mechanical or lightenergy before being converted back into electrical energy in the computer. The same andother problems are also encountered when signals are transmitted in the reverse directionfrom computer to human.Substantial research is now ongoing in which biological signals are collected in a moredirect way [2]. Unfortunately, because signals are measured externally to the humanbody, errors are induced because of the necessary energy transformations, overall neuralsignal averaging and the indirect nature of the measurements taken [3].One line of present research involves the development of, what are described as,wearable computers [4]. These devices range from kitting out everyday implements suchas shoes and jewellery with microcomputers, to directly monitoring indications of stressand alertness, and even to fitting a pair of glasses with a mini-television, allowing thewearer a remote televisual input. In reality all of these devices are far removed from thediscussion in this paper, in that they are positioned externally to the human body, theyoperate on signals monitored on the outside of the body and returning signals need to beconverted by the normal/usual human senses.In order to investigate the practicalities of a more direct interface with computers, variousanimal studies have been carried out. Of relevance is the use of an extracted lamprey1

brain to control the movement of a small robot vehicle [5]. The lamprey is appears, hasan innate response to light, which helps it to navigate in water, in particular to stay theright way up. This innate feature has been used to control the robot such that it alwaysmoves in the direction of a bright light source. Although it does function well, the trial ismore a case of proving a point, as; unfortunately, the isolated lamprey brain only has alimited life span of, at most a few hours or so.Research employing rats is also relevant and has been ongoing for some time. JohnChapin chronically implanted electrodes into the motor cortex of rat’s brains [6] in orderto decode neuronal activity immediately prior to a lever pulling exercise that the rateswere trained for. Signals were processed, via the implant, before the rates actually pulledthe lever and a reward was instantly given. Several of the rates learnt that they no longerhad to actually pull the lever, but merely thinking about the action was sufficient toobtain the reward.Of particular relevance to this article however, is the research of Philip Kennedy [7]whereby he enabled a brainstem stroke victim to control the movement of a cursor on acomputer screen. Functional MRI scans were taken to decide on the best positioning forthe electrode, since the patient had been asked to think about moving his hand. Theelectrode, which was positioned where neuronal activity was most pronounced, consistedof a hollow glass cone containing two gold wires. Meanwhile positioned on the outsideof the patient’s head was an amplifier and radio transmitter. Subsequently, when thepatient thought about moving his hand, so associated signals were transmitted to thecomputer to cause the cursor to move around, thereby enabling a basic form of communication.Whilst it is certainly true that much work has also gone in to implant technology for suchas retina and cochlea implants, it is felt that, although there may be some concepts to belearned from, in general the links with this study are not strong. Essentially we wish toinvestigate here an adaptable bi-directional link involving the computer. Such“replacement” implants do not exhibit such characteristics.With the background of other research as described here, the main aim of the researchdiscussed in this paper, was to form an operational adaptive bi-directional link betweenthe human nervous system and a computer. To this end the first named author (KW)volunteered to receive an implant in his median nerve, the main group of nerve fibresrunning from the brain to the hand area. Details of the implant itself, and the operation,will now be given.

Microelectrode Implant

An array of one hundred individual needle electrodes was fired into the median nerve inthe left arm of KW, just below the wrist, on March 14, 2002, at the Radcliffe Infirmary,Oxford, UK. The array measured 4mm x 4 mm with each of the, uniform length, pinsbeing 1.55mm. The median nerve fascicle was approx 4 mm in diameter, hence the pins,when inserted, extended approx 40% of the way into the fascicle.2

One incision, extending for 4 cm proximally, was made just below the wrist, with asecond incision, extending for 2 cm, positioned centrally, 16 cm proximal to the wrist.An open tube was passed between the incisions, following a tunnelling procedure. Thearray, with associated wires, was then passed down the tube from the second incision tothe first. Subsequently the tube was removed, leaving the array adjacent to the mediannerve, with wires running subcutaneously up the arm, exiting at the second incisionwhere they linked to an external electrical connector pad. The array was pneumaticallyinserted into the radial side of the median nerve bundle.

Bi-directional Signalling

By finger movements, neural signals, associated with the muscle contraction, could begenerated and either passed directly to the computer, via an interface unit, or by means of a digital radio link, positioned on a gauntlet arrangement work externally [8]. Bymonitoring and processing these signals, in conjunction with a proportional-in-timecontroller, the movement, in terms of the associated neural signals, could be used tocontrol a number of real-world devices.A constant current stimulator was employed to stimulate the subpopulation of axonsaffected by the electrodes. Charge balanced, bi-phasic rectangular pulses of 200 usecduration were found to be best suited in allowing KW a stimulated sensation, henceenabling feedback from external mechanisms. Currents of below 80 uA in magnitudewere seen to produce little effect, however at that current magnitude many of theelectrodes produced a recordable effect.Stimulation of this nature was first attempted six weeks after implantation. With thestimulation applied randomly, initial trials produced a correct identification over therange 68-73% of the time. Over the period of experimentation this response graduallyimproved. Final test, of the same nature, were carried out immediately prior to extractionof the implant, on June 18

2002. At this time correct identification over the range 93-98% was witnessed. This indicates that KW’s brain gradually learned to recognise theinjected current signals more clearly.

Articulated Hand

Prosthetic hands are conventionally controlled by means of visual feedback to theoperator, with an analogue input channel supplying a control signal to the hand generatedfrom the electromyograms (EMGs) from forearm muscles. It is usual for them to bedirect controlled in both opening and closing, hence ‘Voluntary Opening VoluntaryClosing’ (VOVC). Essentially, if the user can see the articulated hand’s actions and theycan generate the appropriate muscular commands, they can operate a hand effectively.However without any internal intelligence or if the target object is obscured they maycontinue to close until the motor stalls, thus the prostheses cannot be easily controlledwith finesse.3